What If It’s Still Early? The Bull Case for AI Capex, Broadening Earnings, and a Market That May Have More Room to Run

On one side, investors see explosive semiconductor moves, huge AI infrastructure spending, rich headline valuations, sticky inflation, and a coming wave of mega-IPOs. On the other side, earnings keep improving, capital spending is broadening beyond hyperscalers, margins remain resilient, and the rest of the market outside the biggest names may only be in the early stages of recovery.

The most important question is not whether parts of the market look wild. Some parts always do in a bull market. The real question is simpler:

What if this cycle is still early?

Below is the framework that matters most for investors right now: AI capex versus bubble fears, why valuation compression may actually be bullish, what semiconductor earnings are saying, why the S&P 493 may be the most underappreciated story in the market, and how to think about inflation, the Fed, and the coming IPO supply wave.

1.Quick summary: the core bull case

• AI-driven capital spending is broadening across the economy, not staying confined to a handful of mega-cap tech companies.
• Current capex levels do not look like the 2000 bubble when viewed against sales and free cash flow in aggregate.
• Multiple compression this year happened alongside rising earnings estimates, which is a healthier setup than a price-only melt-up.
• Semiconductors have surged, but much of that move has tracked extraordinary earnings growth rather than pure speculation.
• The biggest opportunity may be outside the Mag 7, with equal-weight and smaller-company earnings only beginning to recover after a multi-year slump.
• Inflation is not ideal, but for equities the historical sweet spot has often been moderate inflation rather than ultra-low inflation.
• Small Fed hikes linked to growth have historically been less damaging to stocks than investors assume.
• Even a major IPO wave from companies like SpaceX, OpenAI, and Anthropic may create more short-term absorption issues than long-term structural damage.

2.The first mistake investors are making: treating every capex wave like a bubble

Many investors see one thing and jump straight to the conclusion.

Capital expenditures are rising fast, especially in technology. Therefore, it must be a bubble.

That logic is lazy.

Capex booms can be dangerous, but they are not automatically dangerous. In fact, from a market, earnings, GDP, and jobs perspective, a capex recovery is usually the healthier path. It means corporate America is investing for growth instead of relying on financial engineering.

That distinction matters.

In the decade following the financial crisis, critics spent years arguing that earnings growth was too dependent on buybacks and dividends. Companies were shrinking their float, protecting margins, and avoiding major reinvestment. The complaint back then was that the earnings were lower quality because they were not being driven by real business expansion.

Now the complaint has flipped. Companies are spending heavily again, and some of the same critics still dislike the rally. The objection has just changed form. Before, it was “too many buybacks.” Now it is “too much capex.”

That tells you something important: for a subset of market observers, the issue is not the reason. The issue is that they are predisposed to distrust the rally.

What the capex data actually suggests

When capex is measured relative to sales, spending is accelerating across all 11 sectors, not just in big tech. That is a broadening recovery signal.

When capex is measured relative to free cash flow, the aggregate picture looks even less bubble-like. At the peak of the dot-com bubble, corporate America was spending roughly 3.5 to 4 times free cash flow. Today, even with the hyperscalers included, aggregate spending remains below 1 times free cash flow.

That is a massive difference.

There may be individual names where spending proves excessive. That is always possible. But systemically, the setup does not resemble the excesses of 2000.

3.Why valuation compression could be bullish, not bearish

One of the strangest dynamics this year has been the combination of rising earnings estimates and falling or compressing valuation multiples.

That combination is rare, and it deserves more attention.

Many investors entered the year assuming two things:

• stocks had already gone up too much over the prior three years
• valuations were too expensive to support another leg higher

But quantitatively, those arguments are weaker than they sound.

Momentum often leads to more momentum. Expensive markets do not automatically underperform just because they are expensive. And when the market compresses multiples while earnings rise, that often means skepticism is still doing the work of restraining exuberance.

That is healthy.

The market is not behaving like a euphoric crowd throwing price discipline out the window. It is behaving like a market that still does not fully trust the good news.

That mistrust matters because bull markets often climb a wall of worry. When fear remains persistent, upside can extend longer than most people expect.

All-time highs do not mean there is no fear

This point trips people up constantly.

Investors look at the index near highs and assume confidence must be everywhere. But positioning, volatility behavior, and options activity often tell a different story. Put buying spikes quickly. Fear indicators have remained elevated at different points. Equity investors still appear more anxious than complacent.

That contradiction is not unusual. A market can rise while skepticism stays high. In fact, that is often the best kind of rally because it means there is still doubt to be converted into buying power later.

4.Goldman’s S&P 8000 logic: aggressive, but not irrational

A major bullish argument making the rounds is that AI-driven earnings growth can push the S&P 500 materially higher without requiring a full-blown valuation bubble.

The basic case runs like this:

• AI infrastructure spending is driving real earnings growth
• earnings estimates have moved materially higher after a strong reporting season
• uncertainty around AI and the macro backdrop keeps valuations from expanding too far too fast

That framework is internally coherent.

If earnings are doing the heavy lifting while multiples stay somewhat constrained by skepticism, the market can advance without repeating the excesses of prior bubbles. That is exactly the kind of setup many investors miss because they are waiting for the advance to look emotionally comfortable before they believe it.

5.Big Tech may be becoming “modern industrials”

This is one of the more interesting ways to frame the current tech leadership.

For years, many large technology companies were celebrated as capital-light software machines with extraordinary margins. Their businesses required some R&D and modest capital investment, but not the type of spending burden associated with industrial companies.

That is changing.

The biggest AI winners are now spending at levels that look much more infrastructure-heavy. Data centers, chips, networking, power, cooling, and long-duration buildouts all pull these businesses closer to an industrial profile.

That does not automatically mean they deserve industrial multiples. Their margins are still far better than traditional industrial firms, and so far many have managed to preserve attractive economics even as capex rises. But it does mean investors need to update their framework.

These are no longer purely capital-light franchises.

They are increasingly modern industrials with software economics layered on top.

Why that does not automatically break the bull case

The key issue is not whether capex is rising. It is whether that spending is tied to earnings and future revenue.

During the dot-com era, huge sums were spent with very little evidence of durable margins or profits across the median company. Today, the largest spenders are already highly profitable and funding their investment directly from strong cash generation. Even when the spending is aggressive, it is not being financed the same way as the equity-fueled excesses of 2000.

That distinction is fundamental.

6.The semiconductor surge: bubble behavior or earnings catching up?

This is the area where the market looks the craziest.

Semiconductors have gone parabolic. Some names have exploded higher in very short periods. Memory stocks in particular have made moves that naturally trigger dot-com comparisons.

The instinctive reaction is obvious: this has to be a bubble.

But that conclusion ignores the most important variable: earnings.

What history says about huge industry gains

When an industry is up 70% to 100% over a year, it feels unsustainable. Yet historically, industry groups that produce that kind of move have often continued to outperform over the following year. The hit rate is far from perfect, but it is much better than most investors assume.

Why?

Because these moves are often linked to earnings anomalies, not just speculative excess.

In semiconductors, price appreciation over the last two years has actually lagged the pace of earnings growth for the industry. That is the anomaly. Investors are staring at the price chart and missing that the earnings curve may be even more extraordinary.

That does not eliminate the risk of violent drawdowns. Any stock or industry that rises that fast can fall hard. But it does mean the simple bubble label is incomplete.

Why semiconductor valuations matter here

Another critical point is valuation relative to history.

When semis are in the cheapest valuation bucket relative to their own history, forward returns have been materially stronger. That suggests some of what looks absurd on the surface may still contain real valuation support underneath.

In plain English: a stock or industry can be up massively and still not be expensive relative to the earnings power the market expects.

That is hard for most people to process because humans anchor to price, not earnings trajectories.

7.Why this does not cleanly map to the 2000 tech bubble

The 2000 comparison is the easiest analogy available, which is exactly why it gets overused.

Yes, there are echoes of that era. Semiconductor moves are dramatic. IPO speculation is returning. AI enthusiasm is everywhere. But the internals are different.

The median tech stock is the key difference

In the late 1990s bubble, the median technology stock had already seen earnings peak years before the final blowoff. That was the real disconnect. Price kept running while the fundamental center of gravity had already broken down.

Today, the median tech stock is profitable. The cap-weighted leaders are extremely profitable. Operating margins are still rising. The earnings deterioration that characterized the bubble period is not the dominant feature of the current market.

Could this eventually become a bubble? Of course. Any long advance can become one.

But the setup today is still a long way from the late-stage deterioration that defined 2000.

Memory chips are not fiber optic cable sitting unused

Another weak analogy is the old telecom buildout, where massive infrastructure was installed ahead of demand and then sat underutilized for years.

That is not a perfect comparison for AI memory and compute. Much of what is being sold now is being put to use immediately because demand already exists. The issue is not whether customers might someday need it. In many cases, the constraint is that current demand already exceeds available compute.

That does not make every valuation safe. It does make the old telecom analogy less precise than many headlines suggest.

8.The hidden story of this market: the S&P 493 may just be getting started

If there is one theme that looks deeply underappreciated, this is it.

Most of the focus remains on the Mag 7 and AI infrastructure winners. Fair enough. They have earned the attention. But the more interesting market development may be what is happening outside that group.

The earnings recovery is broadening.

Why equal-weight earnings matter

For the last three years, cap-weighted earnings growth has been solid, largely driven by the biggest companies. Equal-weight earnings growth has been much weaker and in some cases contractionary for an unusually long period.

This matters because it means a huge part of corporate America spent years in an earnings slump even while the headline index looked healthy.

That slump may now be ending.

The pattern described here is important:

• roughly three years of pressure in equal-weight and smaller-company earnings
• only a few months into a meaningful recovery
• potential symmetry if the recovery phase lasts much longer than investors expect

That is the “what if it’s still early?” case in its clearest form.

If the rest of the market is only four months into a broader earnings rebound after a three-year downturn, then many investors are treating a middle phase like a late phase.

That is a major analytical error.

The market is less dependent on the Mag 7 than the narrative suggests

Another overlooked point is that the Mag 7 have not been the only engine of returns. In fact, at points they have underperformed the broader market. Contributions to index gains have come from a wider set of names, even if many still sit in the broader AI and semiconductor ecosystem.

That is what a broadening bull market looks like: leadership rotates, participation expands, and returns become less concentrated than the headlines imply.

9.Margins outside tech: why profitability may keep surprising to the upside

One of the more controversial bullish calls is that profit margins may remain stronger than many expect, even outside technology.

The reason comes down to one of the most important but under-discussed variables in the economy: unit labor costs.

The margin driver most investors ignore

Over time, lower unit labor costs have had a strong relationship with higher corporate profit margins. If wage pressure is manageable and productivity is improving, companies can protect or even expand margins.

That appears to be part of what is happening now.

Productivity has improved. Wage pressure has not spiraled. Unit labor costs have remained historically favorable. That supports the argument that margin expectations, while optimistic, are not automatically unreasonable.

Some investors assume AI is already the primary reason margins are improving outside tech. That may be premature. The broader margin story still looks more connected to labor productivity and cost structure than to a fully diffused AI productivity boom across every industry.

AI may amplify this later. It does not need to be the full explanation today.

10.Small caps and the manufacturing recovery

Small-cap earnings estimates, particularly in some sectors, look startlingly strong on paper. Some of that can be distorted by the mechanics of moving from losses to modest profits, especially in indexes with a meaningful share of non-earners.

But directionally, the story still matters.

The manufacturing side of the economy has been in a prolonged malaise. New orders and core capital goods trends are now improving. That matters because smaller businesses are more tightly linked to the domestic production cycle.

If manufacturing is finally emerging from a three-year soft patch, then small-cap and equal-weight earnings recovery could be in the early innings too.

That is not a guarantee of immediate outperformance. It is a strong argument that the earnings breadth story is real.

11.Inflation and the Fed: what actually matters for stocks

Investors spend enormous time debating every inflation print and every Fed probability update. Some of that is useful. A lot of it is noise.

The first key point is straightforward: for equities, the ideal inflation regime historically has often been moderate inflation, roughly in the middle range, not ultra-low inflation and not runaway inflation.

In other words, stocks do not require pristine inflation to work.

The danger zone is different than most headlines imply

Historically, the more dangerous setup for equities has been top-quartile inflation that is still rising. A moderate inflation range, by contrast, has often coincided with decent equity outcomes because earnings growth remains the dominant driver.

That is a critical distinction. Investors often talk as if any inflation persistence is automatically bearish for stocks. History is not that binary.

Shelter is distorting the inflation picture

Another important nuance is that core CPI excluding shelter has looked much cooler than headline narratives imply. Shelter remains a major drag in the reported data, but part of that appears tied to a supply constraint dynamic. Higher rates locked many homeowners into existing mortgages, reducing turnover and tightening available supply.

Consumers feel shelter inflation intensely. That is real. But from a policy perspective, it complicates how much the Fed should respond to inflation that is partly shaped by supply mechanics rather than broad overheating.

Oil shocks often make Fed hikes less likely, not more

This is one of the most useful historical observations in the current macro debate.

Many investors assume higher oil means higher inflation, which means a greater chance of Fed tightening. But historically, oil-driven shocks have often reduced the odds of rate hikes because they behave more like a tax on the consumer than a signal of broad overheating.

If gasoline gets more expensive, households have less room to spend elsewhere. That weakens demand in other categories and can limit pass-through effects across the economy.

Raising rates into that type of supply shock can make little sense. It does not fix the price of oil and can worsen the broader economic drag.

What does push the Fed toward hiking? Growth.

The cleaner pattern historically is not oil. It is growth.

When durable goods orders and capital spending are strong, the odds of Fed tightening rise. That is because the economy is showing demand strength, not because energy is temporarily spiking.

Even then, magnitude matters.

12.Why small Fed hikes are not necessarily bearish for equities

This point gets lost constantly in media coverage.

Investors tend to think in binary terms:

• cuts are good
• hikes are bad

That is too simplistic.

Historically, when the Fed is hiking only modestly, equities have often done well. Why? Because small hikes usually reflect an economy that is growing well enough to absorb them. In that environment, rates confirm the cycle rather than derail it.

What really hurts stocks is not a tiny amount of tightening. It is aggressive tightening, deteriorating earnings, or inflation moving into a truly hostile range.

So if the market starts discussing a mild hike instead of an immediate cut, that is not automatically bearish. It depends on why the Fed is moving.

13.The 10-year Treasury: probably less alarming than the headlines suggest

Long-end yields have made investors uneasy at several points. But when viewed relative to nominal growth and the broader economic backdrop, the move does not look especially anomalous.

Lower yields are not always a gift. Often they reflect weaker growth. If the economy is still expanding at a decent pace, somewhat higher long rates can be consistent with that reality.

Again, context matters more than the headline number.

14.The coming IPO wave: SpaceX, OpenAI, Anthropic, and the supply question

This is one of the most interesting tactical issues ahead.

If the market gets a wave of gigantic offerings from companies like SpaceX, OpenAI, and Anthropic, a lot of money will need to be absorbed. Portfolio managers benchmarked to major indexes may effectively be forced to participate to some degree.

That creates a valid question:

Can the market absorb that much new supply without creating disruption elsewhere?

Short term, yes, there can be pressure

On a shorter horizon, the answer is obvious. Large offerings can create rebalancing stress. Investors sell existing winners to fund participation in new issues. There can be temporary volatility and awkward money flows.

That part is real.

Long term, earnings still decide

On a one-year horizon, the argument is more straightforward. If the earnings backdrop remains solid, the market can absorb new issuance. Buybacks can help. New demand can emerge. Initial dislocations do not have to become structural problems.

So yes, mega-IPOs could create short-term turbulence. But long-term damage is far less likely if earnings continue to support the broader market.

15.Speculation is back, but that does not mean the whole market is drunk

There is clearly some froth in the system.

High-beta technology, meme-style baskets, and lower-quality speculative names have all shown signs of life. That is normal in a bull market. There will always be pockets of mania when risk appetite rises.

The key is whether that froth is dominating the entire market or just living on the edges.

The evidence here suggests more of a sideshow than a systemic problem.

When risky tech is normalized against broader tech, the move looks less abnormal than headline anecdotes imply. Even if the junkier part of tech underperforms later, that does not necessarily mean the whole sector or market is in trouble. It may simply reflect the usual pattern of lower-quality names rallying early in a recovery and then lagging more durable businesses over time.

16.The three biggest takeaways for investors

1. We may be much earlier in the broad earnings recovery than the market narrative assumes.**

** The S&P 493, equal-weight earnings, and parts of small-cap corporate America may only be in the first few months of a meaningful rebound after roughly three years of pressure.

1. Multiple compression has been constructive, not destructive.**

** Prices have not outrun fundamentals in the way many feared. Rising earnings combined with skepticism and restrained multiples is a healthier setup than a speculative valuation blowoff.

1. The Fed may matter less than earnings in the second half of the year.**

** Moderate inflation, resilient growth, and even the possibility of small hikes are not necessarily hostile for equities. Earnings breadth and margin durability are likely more important drivers.

17.Bottom line

The cleanest bearish argument right now is easy to understand: AI spending looks huge, semiconductors look stretched, inflation is sticky, and speculative behavior is visible.

The stronger analytical question is whether those observations are enough to conclude the cycle is late.

That is where the data pushes back.

Capex is broadening, not just exploding in one corner. Aggregate spending still does not resemble a systemic bubble. Earnings are doing more of the work than price alone. The market remains skeptical enough to avoid the classic euphoria pattern. And most importantly, the parts of the market outside the mega-cap winners may only now be starting a real earnings recovery.

That is what makes this setup so uncomfortable for many investors. It looks mature on the surface because the headlines are loud and the index has risen for a while.

But under the hood, there is a real case that the broader cycle is not mature at all.

What if it’s still early?